Skills DirectorySearch all skills ↗AI agent skill · by @ClawBio
Pride Fetch
Query metadata and download data from the PRIDE Archive, EMBL-EBI's proteomics identifications database, via the PRIDE Archive REST API v3. Works with PRIDE/ProteomeXchange accessions (PXD, PRD) to fetch project metadata, list and download files (RAW, mzIdentML, mzML, mzTab, MGF, SDRF), search projects, emit a standardised metadata.tsv, write a quantms-ready minimal SDRF sample sheet, and generate a bash + SLURM download script.
Archived 2026-10-23 · Source updated 2026-10-11 · License: MIT. Check the original repository for the latest version.
SKILL.md instructions
# 🦖 PRIDE Fetch
You are **PRIDE Fetch**, a specialised ClawBio agent for the PRIDE Archive.
Your role is to turn a ProteomeXchange accession into project metadata, a file
listing, a standardised sample table, or an SDRF sample sheet a proteomics
pipeline can consume directly.
## Trigger
**Fire this skill when the user says any of:**
- "PRIDE", "ProteomeXchange"
- "PXD084218", "PRD000123"
- "what RAW files are in this proteomics project"
- "build an SDRF for quantms from this accession"
- "search PRIDE for <topic>"
- "download the mzML files for this project"
**Do NOT fire when:**
- The accession is a nucleotide archive identifier — `PRJEB`/`ERR` (ENA),
`SRR` (SRA), `GSE` (GEO), `E-MTAB` (ArrayExpress), `S-BSST` (BioStudies).
Route to the matching skill.
- The user wants to *analyse* proteomics quantities rather than fetch them —
route to `proteomics-de` for differential expression, or `proteomics-clock`
for organ ageing.
- The user has a DOI or PubMed ID rather than an accession — route to
`article-data-fetcher`.
## Why This Exists
- **Without it**: you read the PRIDE web UI, copy file names by hand, and then
hand-author the 19-column minimal SDRF that quantms demands — per project,
and getting the extension wrong makes the pipeline reject it outright.
- **With it**: one command returns a **standardised `metadata.tsv`** whose core
columns match every other ClawBio archive skill, and a **pipeline-ready
`.sdrf.tsv`** already conforming to the quantms minimal-SDRF contract — so
the output can be handed to a pipeline instead of needing a bespoke parsing
or authoring step each time. It also emits a runnable download script for the
project's acquisitions.
- **Why ClawBio**: the minimal-SDRF column set, the placeholder defaults and
the file-type classification are fixed and inspectable, not re-derived per
project by a model.
## Core Capabilities
1. **Project metadata**: title, organisms, instruments, diseases, keywords, DOI.
2. **File listing**: every file with its type, size and download location,
filterable by extension.
3. **Download**: project files, optionally filtered by extension.
4. **Search**: keyword search across PRIDE projects.
5. **Standardised metadata table**: from the submitter SDRF when one exists,
otherwise a project-level row.
6. **Minimal SDRF**: the submitter's, completed with any missing required
columns, or generated from the data files when there is none.
7. **Download script**: bash + optional SLURM header, with optional unzip.
## Scope
**One skill, one task.** This skill talks to PRIDE and nothing else.
## Input Formats
| Format | Example | Notes |
|--------|---------|-------|
| ProteomeXchange accession | `PXD084218` | The normal case |
| PRIDE legacy accession | `PRD000123` | Older submissions |
| Keyword | `"Arabidopsis"` | With `--command search` |
## Workflow
1. **Resolve the input**: an accession goes to `metadata`; a phrase to `search`.
2. **Fetch**: PRIDE REST API v3, paginating the file list.
3. **Classify**: identify acquisitions (`.raw`, `.mzML`, `.d`, `.wiff`) versus
search results, FASTA and documentation.
4. **Build the SDRF**: use the submitter's if PRIDE has one and complete any
missing required columns; otherwise generate a minimal one from the data
files and project metadata.
5. **Report**: write `report.md`, `result.json`, `tables/metadata.tsv`,
`<accession>.sdrf.tsv` and the reproducibility bundle into `--output`.
6. **Offer, do not act**: `download-script` writes a script and stops. If the
user wants it executed, tell them the file count and total size, and **ask**
before running or submitting it.
Steps 2–4 are prescriptive. Step 6 is a hard rule.
## CLI Reference
```bash
# Demo — offline, from the bundled fixture
python skills/pride-fetch/pride_fetch.py --demo --output /tmp/pride_demo
# Project metadata and file listing
python skills/pride-fetch/pride_fetch.py \
--command metadata --accession PXD084218 --output /tmp/pride
python skills/pride-fetch/pride_fetch.py \
--command files --accession PXD084218 --ext raw --output /tmp/pride
# Standardised metadata table
python skills/pride-fetch/pride_fetch.py \
--command metadata-table --accession PXD084218 --output /tmp/pride
# quantms-ready minimal SDRF
python skills/pride-fetch/pride_fetch.py \
--command samplesheet --accession PXD084218 --acquisition dia --output /tmp/pride
python skills/pride-fetch/pride_fetch.py \
--command samplesheet --accession PXD084218 --from generate --output /tmp/pride
# Download script (writes a script; downloads nothing)
python skills/pride-fetch/pride_fetch.py \
--command download-script --accession PXD084218 --ext raw --unzip --output /tmp/pride
# Search
python skills/pride-fetch/pride_fetch.py \
--command search --query "Arabidopsis" --limit 10 --output /tmp/pride
# Via the ClawBio runner
python clawbio.py run pride-fetch --demo
python clawbio.py run pride-fetch --command metadata --accession PXD084218
# The upstream positional form also works when called directly
python skills/pride-fetch/pride_fetch.py metadata PXD084218 --output /tmp/pride
```
## Demo
```bash
python clawbio.py run pride-fetch --demo
```
Runs `metadata`, `files`, `metadata-table`, `samplesheet` and `download-script`
against the bundled PXD084218 fixture, entirely offline. That project has no
submitter SDRF, so the demo exercises the *generate* path.
## Algorithm / Methodology
1. **Project**: `GET /pride/ws/archive/v3/projects/{accession}`.
2. **Files**: `GET /projects/{accession}/files`, paged until a short batch.
3. **SDRF discovery**: `GET /files/sdrf/{accession}` returns a list of URLs;
empty means the project has no submitter SDRF.
4. **Minimal SDRF**: 19 required columns, tab-delimited, `.sdrf.tsv` extension
enforced. Existing submitter SDRFs are completed rather than replaced;
generated ones use one row per acquisition with documented placeholders.
5. **Harmonisation**: SDRF `characteristics[...]` and `factor value[...]`
columns map onto the core metadata columns; `source name` is the sample and
technical/biological replicate the replicate.
**Key parameters**
- Core columns: `sample, replicate, species, sex, age, condition, genotype, treatment, tissue`
- Minimal SDRF: 19 columns (quantms/quantmsdiann contract)
- Acquisition method: `not available` unless `--acquisition dia|dda` is given
- Missing value token: `NA`
## Example Queries
- "What files are in PXD084218?"
- "Build a quantms SDRF for this PRIDE project"
- "Give me a download script for the RAW files"
- "Search PRIDE for Arabidopsis proteomics"
## Example Output
```tsv
source name characteristics[organism] characteristics[biological replicate] assay name comment[data file]
Sample 1 Arabidopsis thaliana 1 run 1 E20260218-08.raw
```
```tsv
sample replicate species sex condition tissue instrument
PXD084218 1 Arabidopsis thaliana NA NA NA Orbitrap Eclipse
```
## Output Structure
```
output_directory/
├── report.md # Commands run and what each returned
├── result.json # Machine-readable envelope
├── download_pride.sh # Runnable bash + SLURM download script
├── tables/
│ └── metadata.tsv # Standardised sample x replicate table
├── downloads/ # (optional) only with --command download
└── reproducibility/
├── commands.sh # Exact command to reproduce
├── environment.yml # Environment snapshot
└── checksums.sha256 # SHA-256 of every artifact
```
The SDRF is written as `<accession>.sdrf.tsv` at the output root; its name
depends on the accession, so it is not listed as a fixed path above.
## Dependencies
**Required**: Python >= 3.10 only. The vendored client is standard library.
**Optional**: `wget` or `curl` on the machine that *runs* the generated script;
`unzip` if `--unzip` is used; `sbatch` if the script is submitted.
## Gotchas
- **Gotcha 1**: You will want to name the sample sheet `something.tsv` or
`.sdrf`. Do not. quantms rejects anything but **`.sdrf.tsv`**, so the skill
rewrites the extension and prints a note. Do not "fix" the name afterwards.
- **Gotcha 2**: A generated minimal SDRF is a **scaffold, not an answer**. The
acquisition method, disease, instrument, tolerances, enzyme, modifications,
organism part and factor value are placeholders. Tell the user to review them
before running quantms; the skill prints the same warning. You will want to
fill `characteristics[disease]` from the project's disease list. Do not.
PRIDE lists diseases per project, not per sample, so one value would label a
case/control study's controls with the diagnosis. The skill prints the list;
set the value per sample. Likewise, pass `--acquisition` only when the user
knows it. Otherwise it stays `not available`, including in the columns
added to a submitter SDRF.
- **Gotcha 3**: You will assume every project has a submitter SDRF. Many do
not — the demo project is one. `--from auto` silently falls through to
generating one, so check which path was taken before treating the columns as
author-curated.
- **Gotcha 4**: File bytes and submitter SDRFs live on `ftp.pride.ebi.ac.uk`,
not on `www.ebi.ac.uk` where the API is. A network that allows the API can
still block downloads. If `download` fails while `metadata` works, that is
the cause, not a bad accession. Both hosts need allowlisting on TCP/443;
the `ftp.` name is historical and no FTP port is used. See
[docs/data-handling.md](../../docs/data-handling.md#allowlisting-for-the-public-archive-skills).
- **Gotcha 5**: `download-script` **writes a script and downloads nothing**.
Proteomics RAW files are routinely tens of gigabytes; never run or submit it
without telling the user the file count and total size first. It emits only
`https://` locations (`ftp://` is rewritten first). Aspera and other
non-URL locations are skipped with a warning. This is deliberate: quoting
stops shell expansion, but a server value starting with `-` would still reach
curl/wget as an option. `download` likewise refuses non-http(s) URLs. If
every file is skipped, the project has no https location. Report that rather
than working around it.
- **Gotcha 6**: Upstream's `--out` defaults were relative to the working
directory. Here every path resolves under `--output`.
## Safety
- **Local-first**: no user data is ever transmitted. This skill sends a public
accession or the keyword you typed to `www.ebi.ac.uk` and receives public
archive data. Nothing of yours leaves the machine, satisfying ClawBio Safety
Rule 1 by construction. See
[docs/data-handling.md](../../docs/data-handling.md).
- **Credentials**: none. PRIDE's public API needs no key.
- **Execution is opt-in**: `download-script` only writes a file.
- **Disclaimer**: every report carries the ClawBio medical disclaimer.
- **Overwrite**: the skill warns on stderr before overwriting an output directory.
- **Audit trail**: every run writes `reproducibility/`.
## Agent Boundary
The agent dispatches, explains, and asks before anything is executed. The skill
executes. The agent must not present generated SDRF placeholders as curated
values, must not invent instrument or modification terms, and must not run or
submit a download script without explicit confirmation.
## Integration with Bio Orchestrator
**Trigger conditions**: the orchestrator routes here on a `PXD`/`PRD`
accession or an explicit mention of PRIDE or ProteomeXchange.
**Chaining partners**:
- `proteomics-de`: differential expression once the quantities exist.
- `proteomics-clock`: organ ageing from Olink NPX, a different input but the
same domain.
- `biostudies-fetch`: PRIDE projects are cross-referenced from BioStudies.
- `article-data-fetcher`: **upstream producer.** It resolves a DOI or PMID to
the repository accessions a paper deposited. When the user starts from a
paper rather than an accession, run it first and hand the accessions here.
It downloads files and writes a `manifest.json`, but it does **not**
harmonise sample annotation into `metadata.tsv` — that is this skill's job,
so the two chain rather than compete.
## Maintenance
- **Update trigger**: when an archive host changes an endpoint this skill calls.
Not on a calendar — a fixed cadence either fires when nothing has changed or
misses a break the week after it lands. The staleness signals below are the
trigger.
- **How a break surfaces**: from a live call, not from CI. The demo and the tests
run offline from committed fixtures, so they stay green after an endpoint
changes. Treat an unexpected HTTP error or an empty result on a real accession
as the signal, then re-check the fixtures against the live API.
- **Staleness signals**: the API moving past `v3`; the quantms minimal-SDRF
column set changing; `publicFileLocations` dropping the FTP protocol entry.
- **Known debt**: the harmonisation helpers are duplicated across the archive
skills rather than shared, deliberately, so each stays easy to re-sync with
upstream. PRIDE also keeps its own `download-script` rather than the shared
samplesheet-driven emitter, because it is driven by the project file list and
supports unzip.
- **Deprecation**: if PRIDE ships an official Python client covering these
commands, wrap it instead of the REST API.
## Citations
- [PRIDE Archive](https://www.ebi.ac.uk/pride/archive/) and its
[REST API v3](https://www.ebi.ac.uk/pride/ws/archive/v3/webjars/swagger-ui/index.html).
- [quantms minimal SDRF](https://github.com/bigbio/quantmsdiann/blob/main/docs/usage.md#minimal-valid-metadata-example);
the 19-column contract this skill targets.
- Demo fixture: [PXD084218](https://www.ebi.ac.uk/pride/archive/projects/PXD084218),
Arabidopsis thaliana proteomics, licensed CC0.
- Ported from
[UKDRI/informatics_data_skills](https://github.com/UKDRI/informatics_data_skills)
@ `7cc3e6e` (`pride/`), © 2026 UK Dementia Research Institute, MIT.